Extraction from streams (#25)

Add the ! postfix operator and !! expression.
  Also add the ++ string concatenation operator.
  Also allow specification of standard input in the test scheme.

  Resolves #7, #18.

Co-authored-by: Glen Whitney <glen@studioinfinity.org>
Reviewed-on: #25
Co-Authored-By: Glen Whitney <glen@nobody@nowhere.net>
Co-Committed-By: Glen Whitney <glen@nobody@nowhere.net>

.drone.yml

@@ -29,7 +29,18 @@ steps:
       - java -jar /drone/lib/spt/org.metaborg.spt.cmd/target/org.metaborg.spt.cmd* -l . -s /drone/lib/spt/org.metaborg.meta.lang.spt -t tests
       - mkdir -p lib
       - curl -o lib/sunshine.jar -L 'http://artifacts.metaborg.org/service/local/artifact/maven/redirect?r=snapshots&g=org.metaborg&a=org.metaborg.sunshine2&v=LATEST'
-      - bin/fosgen tests/emit_sum.fos
+  - name: setup_gen
+    image: gcc
+    volumes:
+      - name: m2
+        path: /root/.m2
+    commands:
+      - git clone https://github.com/facebook/nailgun.git
+      - cd nailgun
+      - make
+      - cd ../bin
+      - ln -s ../nailgun/nailgun-client/target/ng .
+      - cd ..
   - name: extract_tests
     image: xonsh/xonsh
     commands:
@@ -41,7 +52,8 @@ steps:
         path: /drone/lib
       - name: m2
         path: /root/.m2
-    commands:
+    commands:  # Note we first make sure that fosgen is working
+      - bin/fosgen -d tests/emit_sum.fos
       - bin/generate_test_code
   - name: python_tests
     image: python:slim
@@ -55,6 +67,13 @@ steps:
     image: haskell
     commands:
       - bin/run_tests runghc hs
+  - name: ocaml_tests
+    image: ocaml/opam
+    commands:
+      - ls -als tests/extracted
+      - opam init
+      - eval $(opam env)
+      - bin/run_tests ocaml ml
 
 volumes:
   - name: lib

.gitignore

@@ -12,10 +12,16 @@
 
 .pydevproject
 
+a.out
+
 *.aterm
 /site
+bin/ng
 tests/extracted/*
 tests/*.js
 tests/*.py
 tests/*.hs
+tests/*.ml
+tests/*.cmi
+tests/*.cmo
 adhoc*

README.md

@@ -8,9 +8,13 @@ dimensions.
 
 So I embarked on this project to see if I could produce as comfortable a
 language as possible to work in, given that I inevitably will be doing a
-bunch of coding. The language will be
-organized around (unary) ++f++unctions, (binary) ++o++perators, and
-(nullary) ++str++eams, hence the name "fostr".
+bunch of coding. The language will be centrally organized around the
+concept of "streams" (somewhat in the spirit of
+[streem](https://github.com/matz/streem) and/or
+[Orc](http://orc.csres.utexas.edu/index.shtml), or to a lesser extent,
+[Sisal-is](https://github.com/parsifal-47/sisal-is)). In fact all higher-type
+entities will be cast in terms of streams, or in slogan form, "++f++unctions
+and (binary) ++o++perators are ++str++eams" (hence the name "fostr").
 
 Other guiding principles:
 
@@ -24,11 +28,9 @@ Other guiding principles:
   the language design from the ground up, it can be kept both effective and
   natural.
 
-* Code uses functions all the time. So needless to say, functions should be
-  first-class entities that are exceptionally easy to create, pass around,
-  etc.
-
-* And true to the name, operators and streams should be just as easy to handle.
+* fostr code uses streams (and their specializations to functions and
+  operators) all the time, so they are first-class entities that are easy
+  to create, pass around, compose, etc.
 
 * Try to keep the constructs available as simple to reason about as possible,
   and practical to use. So side effects are OK, and it should be clear when

bin/extract_tests.xsh

@@ -13,24 +13,35 @@ DESTINATION = 'tests/extracted'
 # Extension for extracted files:
 EXT = 'fos'
 
+# Extension for desired input:
+INP = 'in'
+
 # Extension for expectations:
 EXP = 'expect'
 
 for path in TEST_LIST:
   destdir = pf"{DESTINATION}/{path.stem}"
   mkdir -p @(destdir)
+  chmod ugo+rwx @(destdir)
   contents = path.read_text()
   tests = re.split(r'test\s*(.+?)\s*\[\[.*?\n', contents)[1:]
   testit = iter(tests)
   for name, details in zip(testit, testit):
-    pfm = re.search(r'\n\s*\]\].*?parse\s*fails', details)
-    if pfm: continue # skip examples that don't parse
-    em = re.search(r'\n\s*\]\]', details)
+    pfm  = re.search(r'\n\s*\]\][\s\S]*?parse\s*fails', details)
+    if pfm: continue   # skip examples that don't parse
+    ntfm = re.search(r'\n\s*\]\].*?don.t.test', details)
+    if ntfm: continue  # explicit skip
+    em = re.search(r'\n\]\]', details)
     if not em: continue
-    example = details[:em.start()+1]
+    example = details[:em.start()+1].replace('[[','').replace(']]','')
     expath = destdir / f"{name}.{EXT}"
     expath.write_text(example)
     echo Wrote @(expath)
+    im = re.search(r'/\*\*\s+accepts.*?\n([\s\S]*?)\*\*/', details[em.end():])
+    if im:
+      ipath = destdir / f"{name}.{INP}"
+      ipath.write_text(im[1])
+      echo "  ...and" @(ipath)
     xm = re.search(r'/\*\*\s+writes.*?\n([\s\S]*?)\*\*/', details[em.end():])
     if xm:
       xpath = destdir / f"{name}.{EXP}"

bin/fosgen

@@ -5,6 +5,7 @@ erro() { printf "%s\n" "$*" >&2; }
 
 ##### Set defaults:
 SUPPRESS_ERR=YES
+USE_NAILGUN=YES
 LANGUAGE=Python
 
 ##### Extract command line options:
@@ -14,18 +15,23 @@ do
     -h|--help)
       echo
       echo "Usage:"
-      echo "  fosgen [-d] [-l LANGUAGE] FILE"
+      echo "  fosgen [-d] [-j] [-l LANGUAGE] FILE"
       echo
       echo "Writes to standard output the code generated from the fostr"
       echo "program in FILE, targeting the specified LANGUAGE (which"
       echo "defaults to Python)."
       echo
       echo "The -d option writes diagnostic output to standard error."
+      echo "The -j option uses the Spoofax Sunshine JAR directly, rather"
+      echo "than via nailgun."
       exit
       ;;
     -d)
       SUPPRESS_ERR=''
       ;;
+    -j)
+      USE_NAILGUN=''
+      ;;
     -l)
       shift
       LANGUAGE="$1"
@@ -67,5 +73,17 @@ then
   exec 2>/dev/null
 fi
 
+if [[ $USE_NAILGUN ]]
+then
+  if [[ $SUPPRESS_ERR ]]
+  then
+    $BINDIR/let_sun_shine
+  else
+    $BINDIR/let_sun_shine noisy
+  fi
+  $BINDIR/ng sunshine transform -p $PROJDIR -n $LANGUAGE -i $PROGRAM
+  exit $?
+fi
+
 java -jar $SUNJAR transform -p $PROJDIR -l $PROJDIR -l $MVN_REPO -n $LANGUAGE -i $PROGRAM
 exit $?

bin/generate_test_code

@@ -4,7 +4,7 @@ failed=0
 
 for dir in tests/extracted/*; do
   for file in $dir/*.fos; do
-    for language in Python Javascript Haskell; do
+    for language in Python Javascript Haskell OCaml; do
       echo bin/fosgen -l ${language%.*} $file ...
       bin/fosgen -l $language $file
       if [[ $? -ne 0 ]]; then

bin/let_sun_shine

@@ -0,0 +1,40 @@
+#!/bin/bash
+
+# Helper for fosgen, not intended to be used directly
+# With an argument, print diagnostic output
+
+BINDIR=$(dirname $BASH_SOURCE)
+
+if $BINDIR/ng sunshine --help
+then
+  if [[ $1 ]]
+  then
+    echo "sun already shining."
+  fi
+else
+  if [[ $1 ]]
+  then
+    echo "disperse the clouds."
+  fi
+  SUNJAR="$BINDIR/../lib/sunshine.jar"
+  PROJDIR="$BINDIR/.."
+  if [[ ! $MVN_REPO ]]; then
+    MVN_REPO="$HOME/.m2/repository"
+  fi
+  if [[ ! -d $MVN_REPO ]]; then
+    MVN_REPO="/root/.m2/repository"
+  fi
+  if [[ ! -d $MVN_REPO ]]; then
+    echo "Cannot find your Maven repository. Please set environment variable"
+    echo "MVN_REPO to its full path and re-run."
+    exit 1
+  fi
+  if [[ $1 ]]
+  then 
+    java -jar $SUNJAR server & 
+  else
+    java -jar $SUNJAR server >/dev/null 2>&1 &
+  fi
+  sleep 5
+  $BINDIR/ng sunshine load -l $PROJDIR -l $MVN_REPO
+fi

bin/run_tests

@@ -9,8 +9,14 @@ diffed=0
 for dir in tests/extracted/*; do
   for file in $dir/*.$ext; do
     ((total++))
-    $command $file > $file.out
-    if [[ $? -ne 0 ]]; then
+    if [[ -f ${file%.*}.in ]]; then
+       cat ${file%.*}.in | $command $file > $file.out
+       result=$?
+    else
+       $command $file > $file.out
+       result=$?
+    fi
+    if [[ $result -ne 0 ]]; then
       echo ERROR: $command $file failed.
       ((failed++))
     else

editor/Analysis.esv

@@ -22,3 +22,4 @@ menus
 
     action: "Show pre-analyzed AST" = debug-show-pre-analyzed (source)
     action: "Show analyzed AST"     = debug-show-analyzed
+    action: "Show analyzed type"    = debug-show-type

editor/Generation.esv

@@ -4,3 +4,4 @@ menus
     action: "Python"     = to-python
     action: "Javascript" = to-javascript
     action: "Haskell"    = to-haskell
+    action: "OCaml"      = to-ocaml

editor/Syntax.esv

@@ -20,6 +20,7 @@ menus
     
     action: "Format"          = editor-format (source)
     action: "Show parsed AST" = debug-show-aterm (source)
+    action: "Desugar AST"     = debug-desugar-fostr (source)
 
 views
   

metaborg.yaml

@@ -19,6 +19,7 @@ language:
   sdf:
     pretty-print: fostr
     sdf2table: java
+    jsglr-version: layout-sensitive
     placeholder:
       prefix: "$"
   stratego:

mkdocs.yml

@@ -2,13 +2,14 @@ site_name: fostr language
 nav:
 - README.md
 - tests/basic.md
+- trans/statics.md
 - implementation.md
 
 plugins:
 - search
 - semiliterate:
     ignore_folders: [target, lib]
-    exclude_extensions: ['.o', '.hi']
+    exclude_extensions: ['.o', '.hi', '.cmi', '.cmo']
     extract_standard_markdown:
       terminate: <!-- /md -->
 theme:

signature/TYPE.str

@@ -0,0 +1 @@
+TYPE.stx

signature/TYPE.stx

@@ -0,0 +1,7 @@
+module signature/TYPE
+signature
+  sorts TYPE // semantic type
+  constructors
+    INT    : TYPE
+    STRING : TYPE
+    STREAM : TYPE

statics/util.stx

@@ -0,0 +1,7 @@
+module statics/util
+imports signature/TYPE
+
+rules
+  lastTYPE : list(TYPE) -> TYPE
+  lastTYPE([T]) = T.
+  lastTYPE([U | TS]) = lastTYPE(TS).

syntax/fostr.sdf3

@@ -8,30 +8,57 @@ context-free start-symbols
   
   Start
 
+lexical sorts
+
+  STRING_LITERAL
+
+lexical syntax
+
+  STRING_LITERAL = "'"~[\']*"'"
+
 context-free sorts
 
-  Start Ex
+  Start LineSeq Line OptTermEx TermExLst TermEx Ex
 
 context-free syntax
   
-  Start.TopLevel = prog:Ex {layout(offside prog)}
+  Start.TopLevel   = LineSeq
 
-  Ex.Int      = INT
-  Ex.Stdio    = <stdio>
-  Ex.Sum      = {Ex "+"}+
-  Ex.Receives = [[Ex] << [Ex]] {left}
-  Ex.Enters   = [[Ex] >> [Ex]] {left}
-  Ex          = <(<Ex>)> {bracket}
+  LineSeq          = Line
+  LineSeq.Sequence = sq:Line+            {layout(align-list sq)}
+
+  Line             = OptTermEx
+  Line.ISequence   = TermExLst OptTermEx {layout(0.first.line == 1.first.line)}
+
+  TermExLst.Prior  = TermEx+
+
+  OptTermEx        = ex:Ex               {layout(offside ex)}
+  OptTermEx        = te:TermEx           {layout(offside te)}
+
+  TermEx.Terminate = <<Ex>;>
+
+  Ex.Int       = INT
+  Ex.LitString = STRING_LITERAL
+  Ex.EscString = STRING
+  Ex.Stream    = <stream>
+  Ex.Sum       = <<Ex> +  <Ex>> {left}
+  Ex.Concat    = <<Ex> ++ <Ex>> {left}
+  Ex.Gets      = [[Ex] << [Ex]] {left}
+  Ex.DefGets   = [<<< [Ex]]
+  Ex.To        = [[Ex] >> [Ex]] {left}
+  Ex.DefTo     = [[Ex] >>>]
+  Ex.Emits     = <<Ex>!>
+  Ex.DefEmits  = <!!>
+
+  Ex           = <(<Ex>)> {bracket}
 
 context-free priorities
 
-  Ex.Enters
-  > Ex.Sum
-  > Ex.Receives,
+  Ex.To
+  > Ex.DefTo
+  > {Ex.Sum Ex.Concat}
+  > Ex.DefGets
+  > Ex.Gets,
 
   // prevent cycle: no singletons
-  Ex.Sum <0> .> {Ex "+"}+ = Ex,
-
-  // flat: no Sum immediately in Sum:
-  {Ex "+"}+ = Ex <0> .> Ex.Sum,
-  {Ex "+"}+ = {Ex "+"}+ "+" Ex <2> .> Ex.Sum
+  LineSeq.Sequence <0> .> Line+ = Line

tests/basic.spt

@@ -1,21 +1,37 @@
 module basic
 language fostr
 
+test hw1_type [[
+[[stream]] << [['Hello,	world! ']] << [[3+2]] << ' times.'
+]]
+run get-type on #1 to STREAM()
+run get-type on #2 to STRING()
+run get-type on #3 to INT()
+run get-type to STREAM()
+/** writes
+Hello,	world! 5 times.**/
+
 /** md
 Title: A whirlwind tour of fostr
 
 ## Whirlwind tour
 
-fostr is just in its infancy, so it's not yet even ready for
-Hello, World. The best we can offer now is this little snippet
-that writes the sum of the ASCII codes for 'H', 'W', and '!' to standard output:
+There seems only to be one way to start a tour like this. So here goes:
 ```fostr
 **/
 
-/** md */ test emit_sum [[
-stdio << 72 + 87 + 33
-]]/* **/ parse to TopLevel(Receives(Stdio(),
-                                    Sum([Int("72"), Int("87"), Int("33")])))
+/** md */ test hello_world [[
+<<< 'Hello, world!'
+]] /* **/
+parse to TopLevel(DefGets(LitString("'Hello, world!'")))
+/** writes
+Hello, world!**/
+
+// Prior proto-hello-world, no longer in the tour.
+test emit_sum [[
+stream << 72 + 87 + 33
+]]
+parse to TopLevel(Gets(Stream(), Sum(Sum(Int("72"), Int("87")), Int("33"))))
 /** writes
 192**/
 
@@ -23,7 +39,7 @@ stdio << 72 + 87 + 33
 ```
 
 At the moment, there are only two ways to run a file containing fostr code
-(you can find the above in `tests/emit_sum.fos`). They both start by
+(you can find the above in `tests/hw.fos`). They both start by
 cloning this fostr project. Then, either:
 
 1. Open the project in Eclipse and build it, visit your program file,
@@ -36,58 +52,99 @@ cloning this fostr project. Then, either:
    
 For example, this snippet generates the following Python:
 ```python
-{! ../tests/emit_sum.py extract:
+{! ../tests/hw.py extract:
   start: 'Stdio\s='
 !}
 ```
-(which writes "192" to standard output), or this non-idiomatic, inefficient, but
-working Javascript:
-```javascript
-{! ../tests/emit_sum.js extract:
-  start: '^}'
-!}
+It generates nearly identical code in
+this simple example for Javascript (just with `"Hello, world!"`
+in place of `r'Hello, world!'`), although it generates a different
+preamble defining Stdio for each language. (Currently, Haskell and OCaml
+code generation are also supported.)
+
+There's not much to break down in such a tiny program as this, but let's do
+it. The prefix operator `<<<` could be read as "the default stream receives...",
+and unsurprisingly in a main program the default stream is standard input and
+output. And `'Hello, world!'` is a literal string constant; what you see is
+what you get. The only detail to know is that such constants must occur
+within a single line of your source file. So depending on how you
+ran the program and how closely you looked at its output,
+you may have noticed this program does not write a newline at the end
+of its message. Nothing is ever implicitly sent to a stream. So if you want
+newlines, you should switch to a (double-quoted) string that allows
+the usual array of escape sequences:
+
+```fostr
+**/
+
+/** md */ test hello_esc_world [[
+<<< "Hello,\t\tworld!\n\n"
+]] /* **/
+parse to TopLevel(DefGets(EscString("\"Hello,\t\tworld!\n\n\"")))
+/** writes
+Hello,		world!
+
+**/
+
+/** md
 ```
-In either case, there's also a preamble defining Stdio that's generated.
-(Haskell code generation is also currently supported.)
+(We threw in two of each so you could clearly see them in the output if
+you run this program.)
 
 ### Everything has a value
 
 As mentioned in the [Introduction](../README.md), everything in a fostr
 program (including the entire program itself) is an expression and has
-a value. So what's the value of that expression above? Well, `stdio` is our
-first example of a stream, and for convenience, the value of a stream
-receiving an item is just the stream back again. The `<<` operator is also
-left-associative, so that way we can chain insertions into a stream:
+a value. So what's the value of that expression above? Well, for convenience,
+the value of a stream receiving an item is (generally) just the stream back
+again. That way we can use the general (left-associative)
+`_stream_ << _value_` operator to chain insertions into a stream:
 ```fostr
 **/
 
-/** md */ test emit_twice [[
-stdio << 72 + 87 + 33 << 291
-]]/* **/ parse to TopLevel(
-  Receives(Receives(Stdio(), Sum([Int("72"), Int("87"), Int("33")])),
-           Int("291")))
+/** md */ test state_obvious [[
+<<< 'Two and ' << 2 << ' make ' << 2+2 << ".\n"
+]] /* **/
+parse to TopLevel(
+  Gets(Gets(Gets(Gets(DefGets(LitString("'Two and '")),Int("2")),
+                 LitString("' make '")),Sum(Int("2"),Int("2"))),
+       EscString("\".\n\"")))
+/** writes
+Two and 2 make 4.
+**/
+
+test emit_twice [[
+stream << 72 + 87 + 33 << 291
+]]
+parse to TopLevel(
+  Gets(Gets(Stream(), Sum(Sum(Int("72"), Int("87")), Int("33"))), Int("291")))
 /** writes
 192291**/
 
 /** md
 ```
-Running this program produces a nice palindromic output: "192291".
 
 And because sometimes you want to emphasize the value and propagate that
 instead of the stream, you can also write these expressions "the other way"
-with `>>`; both forms return the first argument:
+with `>>>` for sending to the default stream or `>>` in general; these forms
+(generally) return the value sent, so the following writes "824":
 ```fostr
 **/
 
 /** md */ test enters_twice [[
-(7 + 8 >> stdio + 9) >> stdio
-]]/* **/ parse to TopLevel(
-  Enters(Sum([Int("7"), Enters(Int("8"), Stdio()), Int("9")]), Stdio()))
+(7 + 8 >> stream + 9) >>>
+]] /* **/
+parse to TopLevel(
+  DefTo(Sum(Sum(Int("7"), To(Int("8"), Stream())), Int("9"))))
 /** writes
 824**/
 
 /** md
 ```
+Two things are worth noting here: the default stream can always be referred to
+directly via the identifier `stream`, and the precedences of `<<` and `>>` are
+different so that generally full expressions go to a stream with `<<` but
+just individual terms are sent with `>>`.
 
 ### Layout in fostr
 
@@ -97,12 +154,13 @@ lines are indented from the start of the initial line:
 **/
 
 /** md */ test receive_enter_break [[
-stdio <<
+<<<
   7
-  + 8 >> stdio
+  + 8 >>>
   + 9
-]]/* **/ parse to TopLevel(
-  Receives(Stdio(), Sum([Int("7"), Enters(Int("8"), Stdio()), Int("9")])))
+]] /* **/
+parse to TopLevel(
+  DefGets(Sum(Sum(Int("7"), DefTo(Int("8"))), Int("9"))))
 /** writes
 824**/
 
@@ -112,20 +170,161 @@ stdio <<
 ```fostr
 **/
 
-/** md */ test enter_receive_bad_break [[
-(7 + 8 >> stdio + 9)
->> (stdio << 9 + 2)
-]] /* **/ parse fails
+/** md */ test enter_receive_bad_continuation [[
+(7 + 8 >>> + 9)
+>> (<<< 9 + 2)
+]] /* **/
+parse fails
 
 /* Extra tests not in the tour */
 test enter_receive [[
-(7 + 8 >> stdio + 9) >> (stdio << 9 + 2)
-]]/* **/ parse to TopLevel(
-  Enters(Sum([Int("7"),Enters(Int("8"),Stdio()),Int("9")]),
-         Receives(Stdio(),Sum([Int("9"),Int("2")]))))
+(7 + 8 >> stream + 9) >> (stream << 9 + 2)
+]] /* **/
+parse to TopLevel(
+  To(Sum(Sum(Int("7"),To(Int("8"),Stream())),Int("9")),
+     Gets(Stream(),Sum(Int("9"),Int("2")))))
 /** writes
 81124**/
 
 /** md
 ```
+
+Of course, fostr programs are not limited to one line; expressions on successive
+lines are evaluated in sequence. For example, the program
+```fostr
+**/
+
+/** md */ test emit_thrice [[
+  <<< 72 + 87
+  <<< 88
+    + 96
+  99 + 12
+       >>>
+
+]] /* **/
+parse to TopLevel(Sequence([
+  DefGets(Sum(Int("72"), Int("87"))),
+  DefGets(Sum(Int("88"), Int("96"))),
+  Sum(Int("99"), DefTo(Int("12")))
+]))
+/** writes
+15918412**/
+
+/** md
+```
+
+will write 15918412. The fostr parser enforces that successive expressions
+in sequence align at the left; e.g., the following fails to parse:
+
+```fostr
+**/
+
+/** md */ test emit_thrice_bad_alignment [[
+  <<< 72 + 87
+<<< 88
+    + 96
+  99 + 12 >>>
+]] /* **/
+parse fails
+
+/** md
+```
+
+Note you can optionally terminate an expression in a sequence with a semicolon,
+and you may place multiple expressions on a single line if the earlier one(s)
+are so terminated. So the following is OK:
+
+```fostr
+**/
+
+/** md */ test emit_several [[
+  <<< 1 + 2; 3 >>>
+  (4 + 5) >>>; stream << 6;
+  <<< 7
+  <<< 8
+    + (9+10);
+  11 + 12 >>>; 13 >>>
+                  >>>
+]] /* **/
+parse to TopLevel(Sequence([
+  ISequence(Prior([Terminate(DefGets(Sum(Int("1"), Int("2"))))]),
+            DefTo(Int("3"))),
+  ISequence(Prior([Terminate(DefTo(Sum(Int("4"), Int("5"))))]),
+            Terminate(Gets(Stream(), Int("6")))),
+  DefGets(Int("7")),
+  Terminate(DefGets(Sum(Int("8"), Sum(Int("9"), Int("10"))))),
+  ISequence(Prior([Terminate(Sum(Int("11"), DefTo(Int("12"))))]),
+            DefTo(DefTo(Int("13"))))
+]))
+/** writes
+3396727121313**/
+
+/** md
+```
+**/
+
+test emit_several_desugar [[
+  stream << 1 + 2; 3 >> stream
+  (4 + 5) >> stream; stream << 6;
+  stream << 7
+  stream << 8
+    + (9+10);
+  11 + 12 >> stream; 13 >> stream
+                        >> stream
+]] /* don't test */
+run desugar-fostr to TopLevel(Sequence([
+  Terminate(Gets(Stream(), Sum(Int("1"), Int("2")))),
+  To(Int("3"), Stream()),
+  Terminate(To(Sum(Int("4"), Int("5")), Stream())),
+  Terminate(Gets(Stream(), Int("6"))),
+  Gets(Stream(), Int("7")),
+  Terminate(Gets(Stream(), Sum(Int("8"), Sum(Int("9"), Int("10"))))),
+  Terminate(Sum(Int("11"), To(Int("12"), Stream()))),
+  To(To(Int("13"), Stream()), Stream())
+]))
+
+test emit_several_default [[
+<<< 1 + 2; 3 >>>
+(4 + 5) >>> >> stream; stream << 6;
+<<< 7 << 75
+<<< 8
+  + (9+10);
+11 + 12 >>>; 13 >>>
+                >>>
+]] parse succeeds
+/** writes
+3399677527121313**/
+
+/** md
+### Streams are bidirectional
+
+So far we have only sent items to a stream. But we can extract them from
+streams as well, with the `!` postfix operator. `!!` all by itself abbreviates
+`stream!`, i.e., extraction from the standard stream. For example,
+
+```fostr
+**/
+
+/** md */ test custom_hw [[
+<<< "What is your name?\n"
+<<< 'Hello, ' ++ !!
+]] /* **/
+parse to TopLevel(Sequence([
+  DefGets(EscString("\"What is your name?\n\"")),
+  DefGets(Concat(LitString("'Hello, '"),DefEmits()))
+]))
+/** accepts
+Kilroy
+**/
+/** writes
+What is your name?
+Hello, Kilroy
+**/
+
+/** md
+```
+
+queries users for their name and then writes a customized greeting. It also
+illustrates the use of `++` for string concatenation, as opposed to `+` for
+(numerical) addition.
 **/

tests/emit_several.fos

@@ -0,0 +1,7 @@
+<<< 1 + 2; 3 >>>
+(4 + 5) >>> >> stream; stream << 6;
+<<< 7 << 75
+<<< 8
+  + (9+10);
+11 + 12 >>>; 13 >>>
+                >>>

tests/emit_sum.fos

@@ -1 +1 @@
-stdio << 72 + 87 + 33
+stream << 72 + 87 + 33

tests/emit_thrice.fos

@@ -0,0 +1,5 @@
+  stream << 'Some numbers: '
+  stream << 88
+    + 96
+  99 + 12 >>
+    stream

tests/hw.fos

@@ -0,0 +1 @@
+<<< 'Hello, world!'

tests/hw2.fos

@@ -0,0 +1 @@
+<<< "Hello,\t\tworld!\n\n"

trans/analysis.str

@@ -1,5 +1,4 @@
 module analysis
-
 imports
 
   statixruntime
@@ -9,6 +8,7 @@ imports
   injections/-
 
   libspoofax/term/origin
+  desugar
 
 rules // Analysis
 
@@ -19,7 +19,8 @@ rules // Analysis
   // multi-file analysis
 //  editor-analyze = stx-editor-analyze(pre-analyze, post-analyze|"statics", "projectOk", "fileOk")
 
-  pre-analyze  = origin-track-forced(explicate-injections-fostr-Start)
+  pre-analyze  = desugar-fostr
+                 ; origin-track-forced(explicate-injections-fostr-Start)
   post-analyze = origin-track-forced(implicate-injections-fostr-Start)
 
 rules // Editor Services
@@ -31,16 +32,36 @@ rules // Editor Services
 rules // Debugging
 
   // Prints the abstract syntax ATerm of a selection.
-  debug-show-aterm: (selected, _, _, path, project-path) -> (filename, result)
+  debug-show-aterm: (sel, _, _, path, projp) -> (filename, result)
     with filename := <guarantee-extension(|"aterm")> path
-       ; result   := selected
+       ; result   := sel
+
+  // Prints the desugared abstract syntax ATerm of a selection.
+  debug-desugar-fostr: (sel, _, _, path, projp) -> (filename, result)
+    with filename := <guarantee-extension(|"desugared.aterm")> path
+       ; result   := <desugar-fostr> sel
 
   // Prints the pre-analyzed abstract syntax ATerm of a selection.
-  debug-show-pre-analyzed: (selected, _, _, path, project-path) -> (filename, result)
+  debug-show-pre-analyzed: (sel, _, _, path, projp) -> (filename, result)
     with filename := <guarantee-extension(|"pre-analyzed.aterm")> path
-       ; result   := <pre-analyze> selected
+       ; result   := <pre-analyze> sel
 
   // Prints the analyzed annotated abstract syntax ATerm of a selection.
-  debug-show-analyzed: (selected, _, _, path, project-path) -> (filename, result)
+  debug-show-analyzed: (sel, _, _, path, projp) -> (filename, result)
     with filename := <guarantee-extension(|"analyzed.aterm")> path
-       ; result   := selected
+       ; result   := sel
+
+  // Extract the type assigned to a node by Statix
+  get-type: node -> type
+    where
+      // Assigns variable a to be the result of the Statix analysis of the entire program (or throws an error)
+      a := <stx-get-ast-analysis <+ fail-msg(|$[no analysis on node [<strip-annos;write-to-string> node]])>;
+      // Gets the type of the given node (or throws an error)
+      type := <stx-get-ast-type(|a) <+ fail-msg(|$[no type on node [<strip-annos;write-to-string> node]])> node
+
+  fail-msg(|msg) = err-msg(|$[get-type: [msg]]); fail
+
+  // Prints the analyzed type of a selection.
+  debug-show-type: (sel, _, _, path, projp) -> (filename, result)
+    with filename := <guarantee-extension(|"type.aterm")> path
+       ; result   := <get-type> sel

trans/desugar.str

@@ -0,0 +1,24 @@
+module desugar
+imports libstrategolib signatures/-
+
+rules
+  /* ISequence() and Prior() are just noise for more expressions in sequence,
+     put in to get the layout rules right. So we remove them and collapse
+     all occurrence of them into one big Sequence() call on a list.
+     
+     This is slightly tricky because there might not be any Sequence() call
+     at the top level, but yet an ISequence(). So we do it in two passes,
+     first converting ISequence()s to Sequence()s, and then collapsing
+     Sequence()s.
+  */
+  deISe:      ISequence(Prior(l),x) -> Sequence(<conc>(l, [x]))
+
+  seqFlatten: Sequence(l) -> Sequence(<mapconcat(?Sequence(<id>) <+ ![<id>])>l)
+  
+  defStream:  DefGets(x)  -> Gets(Stream(), x)
+  defStream:  DefTo(x)    -> To(x, Stream())
+  defStream:  DefEmits()  -> Emits(Stream())
+
+strategies
+
+  desugar-fostr = bottomup(try(defStream <+ deISe <+ seqFlatten))

trans/fostr.str

@@ -6,6 +6,7 @@ imports
   pp
   outline
   analysis
+  ocaml
   haskell
   javascript
   python

trans/haskell.str

@@ -1,49 +1,99 @@
 module haskell
-imports libstrategolib signatures/- util
-
-signature
-  constructors
-    TopLevel: Ex -> Ex
-
+imports libstrategolib signatures/- signature/TYPE util analysis
 rules
-  /* Approach: Generate code from the bottom up.
-     At every node, we create a pair of the implementation and
-     necessary preamble of IO actions.
-     We concatenate preambles as we go up.
-     Finally, at the toplevel we emit the preamble before returning the
-     final value.
+  /* Approach:
+     A) We will define a local transformation taking a term with value strings
+        at each child to a value string for the node.
+     B) We will append IO actions needed to set up for the value progressively
+        to a Preactions rule (mapping () to the list of actions). There will
+        be a utility `add-preaction` to append a new clause to value of this
+        rule.
+     C) We will use bottomup-para to traverse the full AST with the
+        transformation from A so that we have access to the original expression
+        (and can get the Statix-associated type when we need to).
+        Hence the transformation in (A) must actually take a pair of
+        an (original) term and a term with value strings at each child,
+        and be certain to return a value string.
+
+     Finally, at the toplevel we emit the result of <Preactions>() before
+     returning the final value.
    */
 
-  hs: TopLevel((c,p)) -> $[import System.IO
+  hs: (_, TopLevel(val)) -> $[-- Preamble from fostr
+  import System.IO
   data IOStream = StdIO
 
-  stdio :: IO IOStream
-  stdio = return StdIO
-
-  receives :: Show b => IO a -> b -> IO a
-  receives s d = do
-    temp <- s
+  -- Danger: These currently assume the stream is StdIO
+  gets :: Show b => a -> b -> IO a
+  gets s d = do
     putStr(show d)
-    return temp
+    return s
+
+  getsStr :: a -> String -> IO a
+  getsStr s d = do
+    putStr(d)
+    return s
+
+  emit s = do
+    l <- getLine
+    return (l ++ "\n")
 
   main = do
-    [p]return [c]]
+    [<Preactions>()]return [val]]
 
-  hs: Stdio() -> ("stdio", "")
-  hs: Int(x) -> (x, "")
-  hs: Sum((c,p)) -> ($[sum [c]], p)
-  hs: Receives((c, p), (d, s)) -> ($[[c] `receives` [d]], <conc-strings>(p,s))
-  hs: Enters((c, p), (d, s)) -> <hsenter>(c,d,<conc-strings>(p,s),<newname>"fos")
+  hs: (_, Stream())     -> "StdIO"
+  hs: (_, Int(x))       -> x
+  hs: (_, LitString(x)) -> <haskLitString>x
+  hs: (_, EscString(x)) -> x
+  hs: (_, Sum(x, y))    -> $[([x] + [y])]
+  hs: (_, Concat(x, y)) -> $[([x] ++ [y])]
 
-  hsenter: (x, s, p, v) -> (v, <concat-strings>[$[[p]let [v] = [x]], "\n",
-                               $[[s] `receives` [v]], "\n"])
+  hs: (Gets(_, xn), Gets(s, x)) -> v
+    with v := <newname>"_fostr_get"
+       ; <add-preactions>[$[[v] <- [<hs_gets>(s, xn, x)]]]
+  hs: (To(xn, _), To(x, s))     -> v
+    with v := <newname>"_fostr_to"
+       ; <add-preactions>[$[let [v] = [x]], <hs_gets>(s, xn, v)]
 
-  hslist: x -> (<map(Fst); join(|", "); brack>x, <map(Snd); concat-strings>x)
-  brack: x -> $<[<x>]>
+  hs_gets: (s, xn, x ) -> $[[s] [<hs_getOp>xn] [x]]
+  hs_getOp = get-type; (?STRING() < !"`getsStr`" + !"`gets`")
+
+  hs: (_, Emits(s)) -> v
+    with v := <newname>"_fostr_emitted"
+       ; <add-preactions>[$[[v] <- emit [s]]]
+
+  hs: (_, Terminate(x)) -> $[[x];;]
+  hs: (_, Sequence(l))  -> <last>l
+  /* One drawback of using paramorphism is we have to handle lists
+     explicitly:
+   */
+  hs: (_, []) -> []
+  hs: (_, [x | xs]) -> [x | xs]
+
+  /* Another drawback of using paramorphism is at the very leaves we have
+     to undouble the tuple:
+   */
+  hs: (x, x) -> x where <is-string>x
+
+  /* Characters we need to escape in Haskell string constants */
+  Hascape: ['\t' | cs ] -> ['\', 't'  | cs ]
+  /* I think I can just use ASCII constants for characters... */
+  Hascape: [ 0   | cs ] -> ['\', '0'  | cs ]
+  Hascape: [ 7   | cs ] -> ['\', 'a'  | cs ] // Alert
+  Hascape: [ 8   | cs ] -> ['\', 'b'  | cs ] // Backspace
+  Hascape: [ 11  | cs ] -> ['\', 'v'  | cs ] // Vertical tab
+  Hascape: [ 12  | cs ] -> ['\', 'f'  | cs ] // Form feed
 
 strategies
+  haskLitString = un-single-quote
+                ; string-as-chars(escape-chars(Escape <+ Hascape))
+                ; double-quote
 
-  haskell = bottomup(try(hs <+ hslist))
+  haskell = rules(Preactions: () -> ""); bottomup-para(try(hs))
+
+  /* See "Approach" at top of file */
+  add-preactions = newp := <conc-strings>(<Preactions>(), <lines>)
+                 ; rules(Preactions: () -> newp)
 
   // Interface haskell code generation with editor services and file system
   to-haskell: (selected, _, _, path, project-path) -> (filename, result)

trans/javascript.str

@@ -1,31 +1,60 @@
 module javascript
 imports libstrategolib signatures/- util
 
-signature
-  constructors
-    TopLevel: Ex -> Ex
-
 rules
-  js: TopLevel(x) -> $[const Stdio = {
-    receives: v => { process.stdout.write(String(v)); return Stdio; },
+  js: TopLevel(x) -> $[// Fostr preamble
+  const _fostr_readline = require('readline');
+  const _fostr_events = require('events');
+  const _fostr_rl = _fostr_readline.createInterface({input: process.stdin});
+  const Stdio = {
+    gets:  v => { process.stdout.write(String(v)); return Stdio; },
+    emit: async () => {
+            const [line] = await _fostr_events.once(_fostr_rl, 'line');
+            return line + "\n"; }
   }
-  function forwards(data, strm) {
-    strm.receives(data);
+  function to(data, strm) {
+    strm.gets(data);
     return data;
   }
-  [x]]
 
-  js: Stdio() -> $[Stdio]
-  js: Int(x) -> x
-  js: Sum(x) -> $[[x].reduce((v,w) => v+w)]
-  js: Receives(x, y) -> $[[x].receives([y])]
-  js: Enters(x, y) -> $[forwards([x],[y])]
+  const _fostr_body = async () => {
+  // End of preamble
 
-  jslist: x -> $<[<<join(|", ")>x>]>
+    [x]
+
+  // Fostr coda
+    _fostr_rl.close()
+  }
+  _fostr_body();
+  ]
+    with line := "[line]"
+
+  js: Stream()     -> $[Stdio]
+  js: Int(x)       -> x
+  js: LitString(x) -> <javaLitString>x
+  js: EscString(x) -> x
+  js: Sum(x, y)    -> $[[x] + [y]]
+  js: Concat(x, y) -> $[[x] + [y]]
+  js: Gets(x, y)   -> $[[x].gets([y])]
+  js: To(x, y)     -> $[to([x],[y])]
+  js: Emits(x)     -> $[(await [x].emit())]
+  js: Terminate(x) -> x
+  js: Sequence(l)  -> <join(|";\n")>l
+
+  /* Characters we need to escape in Javascript string constants */
+  Jscape: ['\t' | cs ] -> ['\', 't'  | cs ]
+  /* I think I can just use ASCII constants for characters... */
+  Jscape: [ 0   | cs ] -> ['\', '0'  | cs ]
+  Jscape: [ 8   | cs ] -> ['\', 'b'  | cs ] // Backspace
+  Jscape: [ 11  | cs ] -> ['\', 'v'  | cs ] // Vertical tab
+  Jscape: [ 12  | cs ] -> ['\', 'f'  | cs ] // Form feed
 
 strategies
+  javaLitString = un-single-quote
+                ; string-as-chars(escape-chars(Escape <+ Jscape))
+                ; single-quote
 
-  javascript = bottomup(try(js <+ jslist))
+  javascript = bottomup(try(js))
 
   // Interface javascript code generation with editor services and file system
   to-javascript: (selected, _, _, path, project-path) -> (filename, result)

trans/ocaml.str

@@ -0,0 +1,66 @@
+module ocaml
+imports libstrategolib signatures/- util signature/TYPE analysis
+
+/* Note will use bottomup-para to traverse the full AST so that
+   we have access to the original expression (and can get the
+   Statix-associated type when we need to).
+   
+   This means that every one of our local rules must take a pair
+   of an original term and a term with every child replaced by
+   its generated code.
+ */
+
+rules
+  ml: (_, TopLevel(x)) -> $[(* fostr preamble *)
+  type stream = { getS: string -> stream; emitS: unit -> string }
+  let rec stdio = {
+    getS = (fun s -> print_string s; stdio);
+    emitS = (fun () -> (read_line ()) ^ "\n");
+  };;
+  (* End of preamble *)
+
+  [x]]
+
+  ml: (_, Stream())     -> $[stdio]
+  ml: (_, Int(x))       -> x
+  ml: (_, LitString(x)) -> $[{|[<un-single-quote>x]|}]
+  ml: (_, EscString(x)) -> x
+  ml: (_, Sum(x, y))    -> $[[x] + [y]]
+  ml: (_, Concat(x, y)) -> $[[x] ^ [y]]
+
+  ml: (Gets(_,yn), Gets(x, y))
+    -> $[([x]).getS ([<ml_str>(yn,y)])]
+  ml: (To(xn,_),   To(x, y))
+    -> $[let _fto = ([x]) in (ignore (([y]).getS ([<ml_str>(xn,"_fto")])); _fto)]
+  ml: (_, Emits(s))     -> $[[s].emitS ()]
+
+  ml: (_, Terminate(x)) -> x
+  ml: (_, Sequence(l))  -> <ml_seq>l
+  
+  ml_seq: [x] -> x
+  ml_seq: [x | xs ] -> $[ignore ([x]);
+[<ml_seq>xs]]
+
+  /* One drawback of using paramorphism is we have to handle lists
+     explicitly:
+   */
+  ml: (_, []) -> []
+  ml: (_, [x | xs]) -> [x | xs]
+
+  /* Another drawback of using paramorphism is at the very leaves we have
+     to undouble the tuple:
+   */
+  ml: (x, x) -> x where <is-string>x
+
+  ml_str: (node, code) -> $[[<ml_string_cast>node]([code])]
+
+strategies
+
+  ml_string_cast = get-type; (?INT() < !"string_of_int" + !"")
+
+  ocaml = bottomup-para(try(ml))
+
+  // Interface ocaml code generation with editor services and file system
+  to-ocaml: (selected, _, _, path, project-path) -> (filename, result)
+    with filename := <guarantee-extension(|"ml")> path
+       ; result   := <ocaml> selected

trans/python.str

@@ -1,34 +1,38 @@
 module python
 imports libstrategolib signatures/- util
-
-signature
-  constructors
-    TopLevel: Ex -> Ex
-
 rules
 
-  py: TopLevel(x) -> $[import sys
+  py: TopLevel(x) -> $[## Fostr preamble
+  import sys
   class StdioC:
-    def receives(self, v):
+    def gets(self, v):
       print(v, file=sys.stdout, end='')
       return self
-  def forwards(data,strm):
-    strm.receives(data)
+    def emit(self):
+      return input() + "\n" # Python inconsistently strips when using input
+  def to(data,strm):
+    strm.gets(data)
     return data
   Stdio = StdioC()
+  ## End of preamble
+
   [x]]
 
-  py: Stdio() -> $[Stdio]
-  py: Int(x) -> x
-  py: Sum(x) -> $[sum([x])]
-  py: Receives(x, y) -> $[[x].receives([y])]
-  py: Enters(x, y) -> $[forwards([x],[y])]
-
-  pylist: x -> $<[<<join(|", ")>x>]>
+  py: Stream()     -> $[Stdio]
+  py: Int(x)       -> x
+  py: LitString(x) -> $[r[x]]
+  py: EscString(x) -> x
+  py: Sum(x,y)     -> $[[x] + [y]]
+  py: Concat(x,y)  -> $[[x] + [y]]
+  py: Gets(x, y)   -> $[[x].gets([y])]
+  py: To(x, y)     -> $[to([x],[y])]
+  py: Emits(x)     -> $[[x].emit()]
+  py: Terminate(x) -> $[[x];]
+  py: Sequence(l)  -> <join(|"\n")>l
 
 strategies
 
-  python = bottomup(try(py <+ pylist))
+  python = bottomup(try(py))
 
   // Interface python code generation with editor services and file system
   to-python: (selected, _, _, path, project-path) -> (filename, result)

trans/statics.stx

@@ -1,14 +1,267 @@
 module statics
 
 imports signatures/fostr-sig
+imports signature/TYPE
+imports statics/util
 
-// see docs/implementation.md for details on how to switch to multi-file analysis
+/** md
+Title: Adding Program Analysis with Statix
+
+## Development of fostr static analysis
+
+This section is more documentation of Spoofax in general and Statix
+in particular than of fostr itself, but is being maintained here in case
+it could be either helpful to someone getting started with Statix or
+helpful in understanding how the static characteristics of fostr were designed.
+
+As mentioned in the [Overview](../README.md), I don't like to program and a
+corollary of that is never to use a facility unless/until there's a need for
+it. So the first few rudimentary passes at fostr simply declared every program
+to be "OK" from the point of view of Statix:
+```statix
+{! "\git docs/statix_start:trans/statics.stx" extract:
+  start: programOk
+  stop: (.*TopLevel.*)
+!}
+```
+
+Then I reached the point at which the grammar was basically just
+```SDF3
+// Start.TopLevel = <Seq>
+// Seq            = <Ex>
+// Seq.Sequence   = sq:Ex+ {layout(align-list sq)}
+// Ex.Terminated  = <<Ex>;>
+{! "\git docs/statix_start:syntax/fostr.sdf3" extract:
+  start: TermEx.Terminate
+  stop: (.*bracket.*)
+!}
+```
+(The first four clauses are in comments because they approximate fostr's
+grammar; it actually uses a few more sorts for sequences of
+expressions, to achieve fostr's exact layout rules. Also note that the parsing
+of literal strings later evolved to include the surrounding single quotes,
+because the rule above implicitly allows layout between the quotes and the
+string contents, creating ambiguity.)
+
+This was the first point at which there were two different types that might
+need to be written to standard output (Int and String), and although of course
+the dynamically-typed Python and Javascript code generated dealt with both fine,
+the Haskell code needed to differ depending on the
+type of the item written (and I hadn't even started OCaml code generation at
+that point since I knew it would be hopeless without statically typing fostr
+programs).
+
+So it was time to bite the bullet and add type checking via Statix to fostr.
+The first step was to replace the simple assertion that any TopLevel
+is OK with a constraint that its Seq must type properly, and an assignment of
+that type to the top level node:
+```statix
+programOk(tl@TopLevel(seq)) :- {T}
+  type_Seq(seq) == T,
+  @tl.type := T.
+```
+Of course, for this to even parse, we must have a definition of `type_Seq`:
+```statix
+{! ../signature/TYPE.stx extract: {start: module, stop: rules} !}
+**/
+
+// see docs/implementation.md for detail on how to switch to multi-file analysis
 
 rules // single-file entry point
 
   programOk : Start
 
-  programOk(TopLevel(_)).
+  /** md
+rules
+  type_Seq : Seq -> TYPE
+```
+  **/
+
+  type_LineSeq : LineSeq -> TYPE
+
+  programOk(tl@TopLevel(seq)) :- {T}
+    type_LineSeq(seq) == T,
+    @tl.type := T.
+
+  /** md
+Now to type a Seq, we look to the syntax, and see that there are two
+possibilities for what it might be: just an Ex, or a Sequence(_) of a
+list of 'Ex's. For the first, Statix does not allow one sort to simply
+"become" another, but the Spoofax infrastructure automatically inserts
+"injection" constructors for us, in this case one named Ex2Seq. So the
+first rule for `type_Seq` is straightforward:
+
+```statix
+  type_Seq(s@Ex2Seq(e)) = T : -
+    type_Ex(e) == T,
+    @s.type := T.
+```
+where of course type_Ex needs its own declaration analogous to the above.
+  **/
+
+  type_Line : Line -> TYPE
+
+  type_LineSeq(ls@Line2LineSeq(l)) = T :-
+    type_Line(l) == T,
+    @ls.type := T.
+
+  /** md
+
+The other (and in fact more typical) rule for `type_Seq`, when it actually
+consists of a sequence of expressions, is a bit more involved. Fortunately
+Statix provides a primitive for mapping over a list, so we can proceed as
+follows:
+```statix
+  types_Exs maps type_Ex(list(*)) = list(*)
+  type_Seq(s@Sequence(l)) = T :- {lt}
+    types_Exs(l) == lt,
+    lastTYPE(lt) == T,
+    @s.type := T.
+```
+Here `lastTYPE` is a function that extracts the last TYPE from a list.
+Unless/until Statix develops some sort of standard library, it must be
+hand-defined, as done in "statics/util.stx" like so:
+```statix
+{! ../statics/util.stx extract: {start: lastTYPE} !}
+```
+  **/
+
+  types_Lines maps type_Line(list(*)) = list(*)
+
+  type_LineSeq(ls@Sequence(l)) = T :- {lt}
+    types_Lines(l) == lt,
+    lastTYPE(lt) == T,
+    @ls.type := T.
+
+  type_OptTermEx : OptTermEx -> TYPE
+
+  type_Line(l@OptTermEx2Line(ote)) = T :-
+    type_OptTermEx(ote) == T,
+    @l.type := T.
+
+  type_Ex     :     Ex -> TYPE
+  type_TermEx : TermEx -> TYPE
+
+  type_OptTermEx(ote@Ex2OptTermEx(e)) = T :-
+    type_Ex(e) == T,
+    @ote.type := T.
+
+  type_OptTermEx(ote@TermEx2OptTermEx(te)) = T :-
+    type_TermEx(te) == T,
+    @ote.type := T.
+
+  /** md
+
+This brings us to the syntax rules for the basic expressions themselves,
+which comprise almost all of the remaining fostr language constructs.
+But first a mechanism suggested by Ivo Wilms to avoid repeating the node
+type annotation in every rule:
+```statix
+  **/
+
+  /** md */
+  ty_Ex : Ex -> TYPE
+
+  type_Ex(e) = ty@ty_Ex(e) :-
+    @e.type := ty.
+  /* **/
+
+  /** md
+```
+At this stage in fostr's development, there was no difference between a
+terminated and unterminated expression, so the typing rule for that
+constructor was trivial:
+```statix
+  ty_Ex(Terminated(e)) = ty_Ex(e).
+```
+  **/
+
+  type_TermEx(te@Terminate(e)) = T :-
+    type_Ex(e) == T,
+    @te.type := T.
+
+  /** md
+
+Now typing literals is straightforward:
+```statix
+{! "\git docs/statix_works:trans/statics.stx" extract:
+  start: '(.*ty_Ex.Int.*\s*)'
+  stop: '/. ../'
+!}
+```
+  **/
+
+  ty_Ex(Int(_))       = INT().
+  ty_Ex(LitString(_)) = STRING().
+  ty_Ex(EscString(_)) = STRING().
+  ty_Ex(e@Stream())   = STREAM().
+
+  /** md
+
+Finally we get to the binary operators, and here we use the pattern found in
+recent versions of the
+"[chicago](https://github.com/MetaBorgCube/statix-sandbox/tree/master/chicago)"
+example language and in the Fall 2020 TU-Delft class lecture on
+[Name Binding and Name Resolution](https://tudelft-cs4200-2020.github.io/lectures/2020/09/24/lecture5/).
+This pattern lets us specify error messages.
+
+```statix
+  **/
+
+  /** md */
+  ty_Ex(Sum(e1, e2)) = INT() :-
+    type_Ex(e1) == INT() | error $[Expression [e1] not an Int in sum.]@e1,
+    type_Ex(e2) == INT() | error $[Expression [e2] not an Int in sum.]@e2.
+
+  ty_Ex(Gets(e1, e2)) = STREAM() :- {T}
+    type_Ex(e1) == STREAM() | error $[Only Streams may receive items.]@e1,
+    type_Ex(e2) == T.
+
+  ty_Ex(To(e1, e2)) = T :-
+    type_Ex(e1) == T,
+    type_Ex(e2) == STREAM() | error $[Items may only be sent to Streams.]@e2.
+  /* **/
+
+  ty_Ex(Concat(e1, e2)) = STRING() :-
+    type_Ex(e1) == STRING() | error $[Expression [e1] not String in concat.]@e1,
+    type_Ex(e2) == STRING() | error $[Expression [e2] not String in concat.]@e2.
+
+  ty_Ex(Emits(e)) = STRING() :- // At the moment, only stream is stdio
+    type_Ex(e) == STREAM() | error $[Only Streams may emit items.]@e.
+
+  /** md
+```
+
+### Using type annotations in transformation
+
+At this point, Statix properly types all of the valid programs of the very
+rudimentary language defined by the grammar above. But the proximate purpose
+for implementing this typing was to aid Haskell code generation. So how
+do we actually use the assigned types in a Stratego transformation?
+
+Statix provides a Stratego api that includes, among other items, strategies
+`stx-get-ast-analysis` and `stx-get-ast-type(|analysis)` that provide access
+to the assigned types. However, it's easiest to use the information via
+a wrapper like this, essentially lifted from the "chicago" language project:
+```stratego
+{! analysis.str extract:
+  start: Extract.the.type
+terminate: Prints.the.analyzed.type
+!}
+```
+
+Now `get_type` run on a node of the analyzed AST produces the assigned `TYPE`
+(as an ATerm in the constructors of sort TYPE in Statix).
+
+Thus, you can select on the assigned type, as in the strategy to select
+the correct Haskell operator to use to send an item to standard output:
+```stratego
+{! haskell.str extract:
+  start: '(.*hs_getOp.=.*)'
+  stop: \s
+!}
+```
+  **/
 
 rules // multi-file entry point
 

trans/util.str

@@ -3,7 +3,7 @@ imports libstrategolib
 
 rules
   join(|infix) : [] -> ""
-  join(|infix) : [x | xs] -> $[[x][<prejoin(|infix)>xs]]
+  join(|infix) : [x | xs] -> <conc-strings>(x, <prejoin(|infix)>xs)
 
   prejoin(|infix) : [] -> ""
-  prejoin(|infix) : [x | xs] -> $[[infix][x][<prejoin(|infix)>xs]]
+  prejoin(|infix) : [x | xs] -> <concat-strings>[infix,x,<prejoin(|infix)>xs]